Father Daddy Capital: A Localization-First Research Framework for Cryptocurrency Quantitative Trading

This project is a localization-first infrastructure for cryptocurrency trading research and paper trading maintained by filevader87 on GitHub. It uses a multi-engine architecture to support swing trading, scalp trading, Polymarket prediction markets, altcoin mining, and full-set arbitrage strategies. It achieves market state-based dynamic capital allocation through a Bayesian calibration layer, neuroplastic network, and meta-controller. Currently, the project is based on paper trading; live trading functionality requires additional configuration to enable. Original link: https://github.com/filevader87/father-daddy-capital, published on May 26, 2026.

The cryptocurrency market features 24/7 uninterrupted trading, high volatility, and low institutional participation, offering unique opportunities for quantitative trading. However, it also faces challenges such as data acquisition and cleaning, strategy backtesting, risk management, and switching between live and paper trading environments. Father Daddy Capital is positioned as a localization-first infrastructure to address these issues, supporting multiple strategies and enabling adaptive optimization.

The project adopts a modular multi-engine design, with each engine focusing on a specific trading style:

1. Swing Trading Engine: Captures medium-term trends (holding positions for days to weeks), combining technical analysis and fundamental factors;
2. Scalp Trading Engine: High-frequency strategy that captures tiny price fluctuations and is latency-sensitive;
3. Polymarket Engine: Involves event-driven strategies, arbitrage opportunities, and liquidity provision;
4. Altcoin Mining Engine: New coin discovery, liquidity mining, cross-chain arbitrage;
5. Full-Set Arbitrage Engine: Triangular arbitrage, cross-exchange arbitrage, spot-futures arbitrage.

The core innovation is the three-layer adaptive architecture:

1. Bayesian Calibration Layer: Establishes prior distributions based on historical data, updates beliefs with new observations, and uses posterior distributions for decision-making, suitable for handling uncertainty;
2. Neuroplastic Network: Dynamically adjusts connection weights, adapts structure, retains effective strategy memory, emphasizing continuous learning;
3. Meta-Controller: Identifies market states (trend/oscillation/high volatility, etc.), selects strategies, dynamically allocates capital, and judges rule conditions, enabling state-based dynamic capital allocation.

The project is implemented in Python with a modular architecture including modules like market_data (standardized data), agents (signal generation), and risk (risk management). It supports three operation modes:

• Paper trading (default): ./scripts/run_local.ps1 -Mode paper (Windows) or ./deploy.sh paper (Linux/macOS);
• Test mode: corresponding parameter is test;
• Live trading: requires configuration of environment variables such as ALPACA_BASE_URL, and must pass deterministic risk gate tests to enable. Security design: Defaults to paper trading mode; live trading will fail and shut down if credentials are missing; keys are not committed to version control; sensitive files are excluded via .gitignore.

Limitations: The project is still in the research phase; live trading needs reinforcement; relies on external APIs (e.g., Alpaca); high volatility risk in the cryptocurrency market; has technical debt (legacy module integration in progress). Industry Significance: Represents the open-source trend of quantitative trading infrastructure, lowering entry barriers; the three-layer adaptive architecture demonstrates the evolution direction of modern quantitative systems from static to dynamic, and from single to multi-strategy collaboration; provides a reference implementation for learners, developers, and researchers.